Industrial Parts Detection Based on Data Correlation Representation
For the realization of automatic identification of industrial parts in automated industrial production,the residual structure of deep residual network is upgraded.The reservoir module is applied to the residual connection structure of the residual network so that each area of the input data can be represented after being correlated with each other.The proposed model is compared with other deep learning models on industrial parts dataset and public dataset.The experimental results show that the proposed residual network with data correlation representation Resnet18-RC is 0.17%,better than ResNet18 on the industrial parts dataset,and the recognition accuracy is higher than other models.The public dataset like CIFAR-10,CIFAR-100 and Tiny-Imagined indicates that the residual network Resnet50-RC is respectively 0.35,0.62,0.54,1.31 per cent,higher than ResNet50 in terms of accuracy,and has good image recognition performanc.
image recognitionresidual neural networkreservoir computingdata correlation representationindustrial parts
万方数据
维普
